University of Groningen
The historical evolution of inequality in Latin America
Frankema, E.H.P.
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to
cite from it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date:
2008
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Frankema, E. H. P. (2008). The historical evolution of inequality in Latin America: a comparative analysis,
1870-2000 Enschede: PrintPartners Ipskamp B.V., Enschede, The Netherlands
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the
author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately
and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the
number of authors shown on this cover page is limited to 10 maximum.
Download date: 14-06-2017
Chapter 4
The Development and Distribution of Mass Education, 18702000: Persistent Inequality or Breaking with History?
4.1 Introduction
In the highly stratified rural societies of colonial and early post-independent Latin America,
education was not regarded as a necessary requirement for a labour force that predominatantly
consisted of subsistence farmers, péones, serfs and slaves. Given the large concentration of
land ownership, the collateral assets needed to invest in education remained beyond reach for
all but a few. Since it was not clear how the rents of public schooling could be appropriated,
the landowning elites were not keen on paying taxes to invest in educational expansion.
Education for the masses was perceived as part of an undesirable process of civil
emancipation undermining the social order and the distributional status quo (Lindert 2004,
Galor and Zeira 1993, Acemoglu and Robinson 2006).
Engerman and Sokoloff (1997, 2000) show that literacy rates in LAC’s lagged behind
those in North America during the entire 19th century and the first decades of the 20th century.
Within Latin America they find literacy rates to be higher in the former colonial periphery
than in the former colonial core. Around the year 1900, Argentina and Uruguay recorded
literacy rates slightly exceeding 50%, which was considerably higher than in Mexico and
Brazil recording 22% and 26% respectively, but much lower than the 83% of Canada in 1861
and the 80% of the USA in 1870. Yet, during the 20th century primary schooling became
universal in nearly all LAC’s and the post-war statistical reports indicate that educational
investments and attainment dramatically increased since 1950 (UNESCO, Statistical
Yearbook, various issues).
With the on-set of modern economic growth the share of agriculture in Latin
American GDP decreased notably. Consequently, the direct contribution of land inequality to
income inequality diminished rapidly. In the wake of a rising demand for human capital by
technology and skill-intensive sectors, the distribution of education became of paramount
importance for the distribution of income. Hence, where land inequality had largely
constrained the opportunities of social mobility in the pre-modern settler colonies, schooling
became the primary determinant of social mobility in the 20th century.
77
The present chapter analyses the long run development and distribution of mass
education in Latin America from 1870 to 2000 in a global comparative perspective. It pays
specific attention to the timing and pace of primary school enrolment expansion. The central
question is to which extent the initial conditions of inequality have affected the development
and distribution of mass education in the long twentieth century and to which extent these
effects were still present at the start of the 21st century. Can we identify clear break points in
the paths of accumulation and distribution? How slow or fast was the spread of mass
education in comparison to other countries? How long did it take before the diffusion of mass
education led to a more egalitarian distribution of schooling years attained? And did the
spread of mass education come along with improvements in the quality of the educational
system or did it go at its expense?
Assessing such comprehensive questions inevitably invokes a certain degree of
subjective judgement and a high degree of generalisation, but a global comparative
framework helps to place some contested “stylized facts” of Latin American educational
development in perspective. Literature interprets the stylized facts of educational progress in
different ways. Most scholars would argue that the unequal distribution of education has
constrained Latin American economic growth and, more generally, can be seen as a crucial
determinant of high income inequality. Some recent studies using the Gini-coefficient of the
attainment distribution, do not find evidence for such a relationship however. This raises the
question whether these studies pick up the effects of recent investment efforts in primary and
secondary education, or that these observations are related to the use of different concepts and
indicators of educational inequality? These diverging views on the extent of educational
inequality in recent years and its impact on income inequality are specifically addressed in
this chapter.
The main conclusions are that the use of different indicators to a large extent explains
the different views on the state and impact of educational inequality in Latin America. In fact,
the increase in primary school enrolment rates was no slower or faster than could be expected
on the basis of the patterns observed in the rest of the world. It has been faster than in the
most advanced industrial countries and it was notably slower than in the poorest developing
countries in Sub Saharan Africa. The expansion of school enrolment came along with a
comparatively egalitarian gender distribution from the late 19th century onwards. Yet, more
than in any other region of the world, the expansion of primary education took place at the
expense of the quality of education. A comparative analysis of the grade enrolment
distribution reveals that it took even the most advanced LAC’s such as Argentina, Chile and
Uruguay at least four decades to achieve acceptable levels of grade promotion and school
completion after having achieved full primary school enrolment rates. Hence, historical
school enrolment rates only make sense in combination with grade enrolment and school
78
completion data. The reduction of educational inequality gained momentum in the 1980’s,
when grade repetition and pre-completion drop out rates were reduced faster than in other
developing regions in the world. So from a long run comparative perspective, the outlook at
present is in many respects much better than a half century ago.
This chapter is structured as follows. Section 4.2 starts with an introduction of the
empirical literature and pays attention to the various definitions of educational inequality
concepts and indicators. Section 4.3 develops the long run perspective and discusses the
diffusion of mass primary schooling using gross enrolment rates for the period 1870-2000.
Section 4.4 shifts attention towards the distribution of attainment in the post-war era and
clarifies why different indicators, such as the Gini-coefficient and the standard deviation of
attainment, reveal such different views on the comparative level of educational inequality in
Latin America. Section 4.5 focuses on the grade enrolment distribution approach. Section 4.6
presents the conclusion.
4.2 Educational inequality in Latin America: different concepts, different indicators, different
views
At face value the concept of “educational inequality” appears rather straightforward:
educational inequality refers to the extent of “variation” around the “average” level of
education, where a larger amount of variation implies a higher level of educational inequality
and vice versa. However, transforming this definition into a workable comprehensive measure
of educational inequality is highly complicated for several reasons. First, the amount of
variation observed depends on the subject categories included, that is, who compares to
whom? Second, what do we mean by the “level” of education? The number of years of
schooling attained, the level of education completed, the quality of education enjoyed or
student performance? How do we measure and compare the qualitative aspects of education?
The limitations of schooling data are well known and basically pre-define what is
meant by “educational inequality” in literature. Differences in knowledge and skills embodied
in persons are in practice approximated by accounting for interpersonal variation in years of
schooling attained. Given the long run perspective adopted in this study, only crude indicators
such as literacy rates and school enrolment rates are available for the entire period 1870-2000.
It is important to keep these limitations in mind, although it will be argued in section 4.5 that
the grade enrolment distribution approach provides more insight in educational quality
differences, which considerably nuances historical comparisons of enrolment and attainment
in the post-war era.
79
A second concern is that the distribution of education cannot be analysed in a
meaningful way without controlling for the “accumulation” or the “average level” of
education. Whenever societies start to broaden their educational basis, increasing differences
in educational experiences inevitably occur. Full primary school enrolment rates cannot be
established overnight and at some point half of a nation’s population will have received at
least some schooling, whereas the other half has not. Given the natural ceiling to the amount
of education each individual can receive, some convergence in the distribution of education is
endogenous to the expansion of education beyond a certain level as well. Hence, comparing
the distribution of attainment levels across countries at a fixed point in time, without
controlling for the stage of educational development, will give results that partly reflect crosscountry differences in educational inequality and for another part (and this can be the major
part) differences in the expansion of education, irrespective of the employed unit of
measurement. Therefore, this study chooses to explicitly link the analysis of educational
development and distribution by focusing on the time it takes societies to achieve certain
benchmark levels of primary school enrolment, grade promotion or school completion. The
idea is that the longer it takes before such benchmark levels are reached, the longer the
negative distributional consequences of a transition towards mass education persist.
Despite these methodological constraints, many scholars find evidence for the conclusion that
educational inequality in Latin America was and still is comparatively large. For instance,
Birdsall and others show in various studies that the accumulation and distribution of
education during the second half of the 20th century in Latin American countries has
developed less favourably than in East Asia. The unequal distribution of education in Latin
America is found to contribute significantly to the region’s modest labour productivity growth
and persistent high levels of income inequality (Bourguignon 1993, Birdsall and Sabot 1994,
Park et al. 1996, Birdsall et al. 1997, Birdsall 1999). This conclusion is based on the analysis
of primary, secondary and tertiary school enrolment and completion rates, the standard
deviation of years of schooling attained, and various measures of educational expenditure.
These indicators reveal, among other things, a bias in public investment towards higher levels
of education combined with relatively poor primary school completion rates in LAC’s.
Birdsall et al. (1997: p. 125) conclude that,
“The unequal distribution of education in Latin America, in terms of both quantity and
quality, constrained economic growth in the region by resulting in forgone opportunities to
increase labor productivity and change household behaviour. At the same time, the relatively
small size of the educated labor force and the resulting high scarcity rents commanded by
educated workers contributed to high inequality in the distribution of income.”
80
Morley (2001) underlines this view arguing that relative wage levels of university graduates
are still higher in LAC’s than in other parts of the world, despite the rapid increase in, and
supply of, university graduates since the 1970’s. Londoño and Székely (2000) find that the
increase in wage differentials between 1982 and 1995 corresponds to increasing skill
differentials across various groups of wage earners. A recent report of Euromonitor
International comparing income distribution across countries lists the ratio of average
disposable income of people who completed tertiary education to the average disposable
income per capita. Figure 4.1 presents the entire sample, subdivided into LAC’s and the rest
of the world in a scatter diagram. The x-axis shows the average per capita disposable income
and the average disposable income of the tertiary educated is at the y-axis. Appendix table
A.4.1 presents the underlying data and lists all the countries included.
Figure 4.1: Average per capita disposable income (x-axis) versus average disposable
income of tertiary educated (y-axis) in 2000 (1995 US $)
50,000
y = 1.52x
40,000
Singapore
30,000
Japan
U.A.E
20,000
y = 2.79x
10,000
0
0
10,000
20,000
Rest of the world
30,000
40,000
50,000
Latin America
Source: Euromonitor International (2007) World Income Distribution 2006/2007, 4th edition, pp. 102-7.
See also appendix table A.4.1.
The estimated linear functions in the figure -with the intercept of both equations set at zeroleave little doubt about the distinctive relation between levels of education and net disposable
income in Latin America. A closer look at the figures in the appendix table shows that the
Latin American average “tertiary education premium” of 251% is only exceeded by three
non-Latin American countries, i.e. Egypt (251.5%), Jordan (260%) and Saudi Arabia (278%).
The average of the rest of the world is 164%. These figures suggest that, either, a) similar
81
skill-differentials in Latin America lead to higher wage differentials than in other parts of the
world, or b) skill-differentials per se are larger than in other parts of the world, or c) both.
However, some recent studies find that the levels of educational inequality in Latin
America are comparatively modest and certainly not way out of line with other countries
(Castello and Domenech 2002, Thomas et al. 2001, Sahn and Younger 2004). These studies
also find that the observable association between educational inequality and income
inequality in Latin America is weak (World Bank 2004). All these studies have one thing in
common: they use the Barro and Lee dataset of educational attainment of the working age
population to calculate Gini-coefficients of the attainment distribution. The Gini, so it is
argued, is a more comprehensive inequality indicator than such “partial” indicators as school
enrolment rates, completion rates or education expenditures per level of education. The recent
World Bank report Inequality in Latin America, Breaking with History? (2004: p. 153)
concludes on the basis of the estimated relationship of educational Gini’s and income Gini’s
that,
“Latin American countries appear to have “too much” income inequality, given their levels of
inequality in years of schooling […] However, before jumping to the conclusion that
educational disparities are definitely not the reason for high income inequality in Latin
America, it should be pointed out that the years of schooling is a very imperfect measure of
the human capital stock embodied in a person”.46
This conclusion is important for two reasons. First, it leaves open the possibility that
educational inequality resides mainly in quality differences rather than in differences in years
of schooling attained. Secondly, the educational Gini apparently leads to other inferences than
the broader set of estimates applied by other studies. It is also noteworthy that Cole et al.
(2004) go a step further claiming that, on the basis of educational attainment data, a lack of
catching up growth of Latin America versus the US can definitely not be explained by a lack
of human capital accumulation. The authors argue that several LAC’s (Argentina, Chile,
Uruguay) obtain equal or higher attainment levels in the labour force in 1990 than many of
the East Asian and European development successes such as Portugal, Spain and Singapore.
Moreover, the average ratio of human capital to output is found to be 40 percent higher in
Latin America than in the USA:
46
This part of the World Bank report is largely based on studies by Castello and Domenech (2002) and
Thomas et al. (2001).
82
“We conclude that human capital is not the major factor in explaining Latin America’s TFP
gap, nor does it appear to play an important role in Latin America’s long run stagnation.”
(Cole et al. 2004: p. 14)
So the question arises whether recent studies are the first to pick up effects of recent changes
in the distribution of education? Does it matter which type of data and indicators were used:
enrolment or attainment data, the Gini-coefficient or the standard deviation? And how
important is the distinction between educational quality as compared to inequality in years of
schooling attained?
4.3 The spread of primary education in Latin America, 1870-2000
Latin American gross primary school enrolment rates for the period 1870-2000 are presented
in appendix table A.4.2. The figures refer to the ratio of enrolled children in the age group 5
to 14 over the country specific primary school age group. The pre-war estimates are retrieved
from Lindert (2004) and Mitchell (2003). The table shows that in the year 2000 all LAC’s
reported gross enrolment rates surpassing 100%, except Haiti.47 The table further shows that
the acceleration in the spread of education in the majority of LAC’s took place in the course
of the 20th century and that the intra-regional dispersion in primary school enrolment rates has
been large until at least the 1970’s. When concentrating on the timing of the transition
towards mass education we can roughly distinguish three groups. In the last three decades of
the 19th century the expansion of primary schooling is most notable in Argentina, Chile, Costa
Rica and Uruguay.48 The British colonies Jamaica and Trinidad & Tobago recorded the fastest
rise and were the only two countries recording a rate exceeding 50% in 1900. After gaining
independency from Colombia in 1903, Panama joined the club of “early movers”. During the
1920’s and 1930’s the rise in gross enrolment rates started to accelerate in Bolivia, the
Dominican Republic, Ecuador, El Salvador, Mexico, Brazil, Peru and Venezuela. Guatemala,
Honduras, Nicaragua and Haiti were typically “late movers”, where the acceleration occurred
only in the early post-war years.
47
Haiti recently stopped reporting enrolment data altogether. Contrary to net enrolment rates, gross
enrolment rates surpass 100% since they are calculated as the ratio of the number enrolled over the
number of children in the specific school age category. For example, a primary education system
containing eight grades the age category is usually 4 to 12. All children enrolled of 13 years or older
are taken into account in the gross enrolment rate, whereas they are excluded in the net enrolment rate.
48
Since a lot of observations for the period 1870-1900 are missing we have to be cautious: a backward
extrapolation of observed trends suggest that the transition towards mass education took place
somewhere between 1870 and 1900 in Costa Rica and Uruguay. Literacy rates recorded in the late 19th
and early 20th century also support the idea that these countries were ahead of the rest of the region
(Thorp 1998, Mariscal and Sokoloff 2000).
83
This classification reflects some important features of Latin America’s colonial
legacy. The “early movers” constitute the countries in the former colonial periphery where the
impact of Iberian metropolitan policies had been markedly smaller than in the core areas such
as New Spain and Peru. These countries further appear to have been a) the most urbanised, b)
the ethnically most homogenous (with larger shares of Europeans or Creoles), c)
comparatively less unequal rural societies (especially Argentina and Costa Rica) and, d) of
British colonial origin (Jamaica and Trinidad and Tobago). The “late movers” are typically
the most stratified and least urbanised rural societies characterised by large ethnic
heterogeneity and a relatively small Creole elite. Yet, the majority of LAC’s fell in between
these extremes and started to invest in mass education in the early 20th century, especially
during the 1920’s and 1930’s.
Figure 4.2: Scatter plot of primary school enrolment rates (age group 5-14) and GDP
per capita (1990 Geary-Khamis US $), Latin America versus Europe, other New World
countries and Japan, 1870-1930
100
80
60
40
20
0
0
1000
2000
3000
Europe, New World, Japan
4000
5000
6000
7000
Latin America
Sources: Maddison 2003, Lindert 2004 and own calculations based on Mitchell 1993. Notes: Countries
included: Argentina, Brazil, Chile, Colombia, Costa Rica, El Salvador, Guatemala, Mexico, Peru,
Uruguay (Latin America); Countries included in the benchmark sample: Austria, Belgium, Canada,
Denmark, Finland, France, Greece, Hungary, Italy, Japan, Netherlands, New Zealand, Norway,
Portugal, Romania, Spain, Sweden, Switzerland, UK, USA.
Figure 4.2 plots gross enrolment rates against levels of GDP per capita for a large sample of
LAC’s and a benchmark sample of economically advanced countries between 1870 and 1930.
It turns out that controlling for GDP per capita, LAC’s had substantially lower enrolment
rates then the control group. The two observations that come close to the trend line of the
84
control group are Costa Rica and Mexico in 1930. Argentina and Chile exceed the Latin
American trend line, while Brazil, Guatemala, Peru and Uruguay are constantly situated
below the trend line. Apparently, LAC’s were, at least from a macro-economic point of view,
“too rich” for their comparative rates of enrolment. Yet, given the historically large inequality
in the distribution of income, assets and wealth among Latin American citizens, the choice for
schooling was in many individual cases hampered by a lack of resources. Hence, it seems
validated to speak of a distinctive “Latin American path” of educational development which
relates to a suboptimal distribution of economic resources.
Early Latin American educational development from a political economic perspective
The “delay” in the transition towards mass education was not so evident during the first years
after the wars of independence in the early 19th century. Influenced by European
enlightenment ideology and the spirit of revolution, education for the masses became an
important topic on the political agenda of the early post-colonial administrations. The decree
of the Peruvian liberator San Martin in 1822 illustrates the revolutionary spirit,
“Public instruction is the primary need of all peoples. Any government that does not promote
it is guilty of a crime which later generations, have the right to avenge, while cursing its
memory.” (UNESCO 1958, 836-7)
In 1825 another famous liberator, Simon Bolivar, ordered the establishment of a teacher
training school in every departmental capital of Peru as part of an ambitious campaign to
implement the revolutionary agenda. Public instruction was thought to be a vital instrument
for the promotion of social, cultural and economic development of the independent Latin
American nation states. The early ideas about public primary education were based on three
leading principles: it should be compulsory, secular and free of charge. However, when postrevolutionary conservative regimes took over, the momentum disappeared as fast as it had
arrived. The three principles not only encountered practical problems but also met with severe
political opposition.
Endemic political instability and chronic budget deficits impeded the required
educational investments and complicated the formation of an efficient bureaucratic apparatus
to initiate and monitor the process. But apart from that, it was the colonial legacy of social
and economic inequality which undermined the sense of urgency among the elite to raise
redistributive taxes to finance public education (Engerman and Sokoloff 2000, Engerman et
al. 2001, Mariscall and Sokoloff 2001). The meagre perspectives of social mobility for the
poor reduced the perceivable benefits of, and consequently, the popular demand for primary
85
schooling. The perceived benefits of schooling were even lower when poor families were
asked to contribute to educational expenses via immediate financial contributions or taxation.
Finally, the principle of secular education met with severe resistance by the Catholic church.
The church perceived public education as one of its traditional domains and feared to loose its
monopoly control over a beloved medium to spread religious ideology and maintain religious
authority (Bakewell 2004).
Hence, it is not surprising that education remained the privilege of a small upper class
during most of the 19th century (Spalding Jr. 1972, Vaughan 1975, Yeager 1991). Education
was deemed important as a means to strengthen national identity, but it was actually used as
an instrument of political control in the hands of the elite (Brock 1985). The World Survey of
Education of the 1950’s (UNESCO 1958) and an ECLAC report one decade later (ECLAC
1968) sum up a long list of problems encountered in the expansion of primary education in
various LAC’s. Among these are 1) a lack of financial resources,49 2) a lack of well educated
teachers, 3) geographical barriers hampering the establishment of schools and school
attendance in isolated rural areas, 4) the language barrier in countries with large indigenous
populations, 5) the indifference towards primary education on behalf of poor and low
educated parents, 6) the practice of child labour provoking irregular school attendance, 7)
insufficient monitoring agencies to detect poor quality and enforce compulsory attendance.
These arguments suggest that a historical legacy of inequality adversely affected the
expansion of mass education in more than one way. There was a lack of political will to
introduce redistributive taxes to invest in schooling for the lower income classes. Yet, the
confined opportunities of social mobility in the highly stratified Latin societies per se, lead to
low perceivable benefits of education among the lower social classes. Hence, institutional
changes were an absolute requirement to break out of this low level equilibrium (Parrado
1998). The conclusions of the Brazilian contribution to the World Survey of Education nicely
illustrate how the poor quality of schooling blends with low perceptions and, consequently, a
persistence of poverty and inequality,
“A school which is not felt to be absolutely necessary, because of its meagre curriculum,
because the basic equipment for life which it gives its pupils is such a poor modicum, must
inevitably be a school to which children only go if they have nothing more important to do.”
(UNESCO 1958, 173)
49
As stated above, this argument does not hold from an aggregate economic viewpoint, but it does
make sense from a distributional point of view.
86
The pace of educational expansion in comparative perspective, 1830-2000
The “delayed” transition towards mass education in Latin America can be explained from its
specific historical heritage. A different question is whether the expansion of primary
education, once underway, was any slower or faster compared to other regions? And were the
“early movers” in Latin America any slower or faster than their neighbours?
Table 4.1 shows the average annual increase of gross primary enrolment rates in a
sample of LAC’s and non-LAC’s from 1830 to 2000 (for the underlying data see appendix
table A.4.1.) The average annual increase of the gross primary enrolment rate refers to the
first observable decade until the decade of “complete” enrolment (labelled average speed) and
in the three consecutive decades with the most rapid expansion observed (labelled maximum
speed).50
Table 4.1 underlines the argument of Clemens (2004) that present-day developing
countries expand school enrolment at a much faster pace than the early industrialising
countries back in the 19th century. In terms of the timing and the rate of expansion LAC’s
mostly fell in between the “early movers” in the industrialising world and the “late movers” in
the poorest parts of the developing world. For instance, with average annual increases
between 0.8 and 2.6 all LAC’s outpaced the USA between 1830 and 1870 (0.6), while Nigeria
(1.9) and Malawi (3.3) were considerably faster than any LAC in the second half of the 20th
century.
Within Latin America the negative correlation between the timing and the pace of
expansion can not be observed. Early movers such as Argentina, Chile and Costa Rica
recorded an average annual increase of 1.1 percent, which equals the Latin American average.
Yet, late movers such as Honduras and Nicaragua achieved average annual increases of 1.4
and 2.1 respectively, which is clearly higher than the regional average. The Dominican
Republic, Venezuela, Peru and El Salvador also show higher rates of expansion during the
mid-twentieth century.
50
It is assumed that countries reporting an enrolment rate of 95% or higher at the start of a particular
decade will achieve a full enrolment rate (100%) in the same decade.
87
Table 4.1: Average annual increase of gross primary enrolment rate, Latin America
versus a selection of non Latin American countries, 1830-2000
Average annual
Three decades of
1870-2000
increase
maximum increase
Argentina
1880-1950
1.1
1890-1920
Bolivia
1900-1990
0.8
1930-1960
1.4
Brazil
1870-1980
0.8
1920-1950
1.2
Chile
1880-1960
1.1
1880-1910
1.8
1.3
Costa Rica
1890-1960
1.1
1890-1920
1.0
Dominican Rep.
1930-1960
1.9
1930-1960
1.9
Ecuador
1920-1970
1.2
1930-1960
1.4
El Salvador
1920-2000
1.1
1930-1960
1.8
Guatemala
1920-2000
0.8
1970-2000
1.5
Honduras
1930-1980
1.4
1940-1970
2.1
Jamaica
1870-1960
0.8
1870-1900
1.2
Mexico
1880-1970
0.8
1920-1950
1.3
Nicaragua
1950-1980
2.1
1950-1980
2.1
Peru
1900-1970
1.3
1920-1950
1.9
Trinidad & Tobago
1870-1960
1.0
1870-1900
1.4
Uruguay
1900-1960
1.3
1930-1960
1.9
Venezuela
1930-1960
2.6
1930-1960
Latin American average
1.25
2.6
1.64
1830-1930
USA
1830-1870
0.6
1830-1870*
0.6
Austria
1840-1920
0.7
1870-1900
0.8
Belgium
1830-1920
0.5
1890-1920
1.1
France
1830-1880
0.9
1850-1880
1.0
Spain
1860-1930
0.6
1900-1930
0.8
UK (England-Wales)
1830-1900
0.7
1830-1860
1.1
Japan
1870-1910
1.1
1870-1900
Average
0.70
1.1
0.92
1930-2000
Philippines
1930-1960
1.6
1930-1960
1.6
Thailand
1930-1980
1.3
1930-1960
1.5
Korea, rep.
1930-1960
2.6
1930-1960
2.6
Turkey
1930-1960
2.5
1930-1960
2.5
Kenya
1930-1980
1.7
1950-1980
2.5
Nigeria
1930-1980
1.9
1950-1980
3.1
Malawi
1970-2000
3.3
1970-2000
Average
2.12
3.3
2.44
Source: Lindert 2004; UNESCO, World Survey of Education, 1958; UNESCO, Statistical Yearbook,
various issues 1966-1998; * USA average over four decades due to lack of intermediate observations.
88
The expansion of mass education from a gender perspective, 1890-2000
Perhaps the most remarkable stylized fact of educational development in the late 19th and
early 20th century relates to the comparatively equal gender distribution of primary school
enrolments in Latin America. Appendix table A.4.3 shows the percentage share of females in
primary school enrolment for the years 1890-1902, 1950-54 and 1990-97. The table also
shows the female shares in secondary and tertiary enrolment for the latter two periods. It turns
out that the female share in primary schooling from the earliest years of the transition
onwards were more or less comparable to those in the most advanced European economies
and the USA and this remained the case throughout the 20th century. The Latin American
average percentage share in the period 1890-1902 is 44.3%.51 Compared to the gender
distribution in European countries such as Portugal and Greece and Asian countries like
Japan, India, Sri Lanka and Myanmar this figure, indeed, is surprisingly high.
Comparatively low levels of gender inequality can also be noted in secondary and
tertiary education. In the 1950’s the average share of female in secondary education was
41.1% and this number increased to 52% in the 1990’s. Although it appears that some of the
Asian countries such as Japan and Sri Lanka had overtaken the Latin American average in the
1950’s, the figure still compares well to such countries as Greece or Spain, let alone the
developing countries in Africa and the Middle East. In tertiary education the figure of 23.8%
in the 1950’s is even higher than in the Netherlands or Switzerland. Yet, it should be noted
that the comparatively high female tertiary enrolment shares are likely to be the result of high
social and economic inequality between Latin American families. The rich elite families are
able to send all their children, boys and girls indifferently, to college or university. In the
Netherlands and Switzerland enrolment in tertiary education was accessible for middle or
lower income groups (with or without public support), but boys were the main beneficiaries
of the increasing public investments in education in first instance. It should be noted that the
comparatively low levels of gender inequality can also be observed in the semi-feudal
societies of Southern Europe, whereas in the Asian countries tertiary education appears to
have been an exclusive male privilege until, at least, the 1950’s.52
In sum, the expansion of mass education was delayed, but once underway it did not move
distinctively slower or faster than could be expected. There are good reasons to interpret the
51
Since this average includes many of the most advanced LAC’s at that time, and excludes most of the
less advanced LAC’s this arguably is an overestimation. Nevertheless, the estimate for Guatemala of
32.8% shows that even in the poorest LAC’s the gender distribution was fairly egalitarian when
compared to all Asian countries observed.
52
The finding of low comparative levels of gender inequality in Latin America is in line with the
results of Camps et al. (2006) who show that gender wage disparities are much lower in Latin America
than in several East Asian countries in the second half of the 20th century.
89
Latin “delay” as the consequence of prevailing social, economic and political inequality
during the 19th century. There were sufficient resources on an aggregate economic level that
could have been devoted to educational expansion, but the resistance to redistributive taxation
in combination with low perceived benefits of education in societies characterised by
confined opportunities of social mobility distorted the required incentives to invest in primary
schooling. These forces obviously lost strength in the course of the 20th century. Given the
historical relationship between economic inequality and the delay in enrolment expansion it is
quite remarkable that the gender distribution of enrolment has been rather egalitarian, even in
comparison to some of the early industrialised countries, but especially in comparison to
Asian, African and Middle Eastern countries.
4.4 The distribution of educational attainment, 1950-2000
This section turns to the question how educational expansion impacted on the distribution of
education. To analyse this issue it focuses on attainment data in the present section and on
grade enrolment data in the next section. A considerable part of current research on
educational inequality is based on educational “stock”, rather than educational “flow” data.
Stock data refer to the years of schooling attained or the level of education completed by the
labour force in a given year (Psacharopoulos and Arriagada 1986, Nehru et al. 1995, Barro
and Lee 2001)53. Attainment figures are widely used for calculating Gini-coefficients
(Castello and Domenech 2002, Thomas et al. 2001, Sahn and Younger 2004, World Bank
2004), standard deviations (Ram 1990, Birdsall 1999) and a measure of the size of the
“educational middle class” such as the percentage share that has completed secondary school
at the highest level of schooling attained (Birdsall et al. 1997).
It turns out that the studies based on Gini-coefficients of the attainment distribution
tend to take a milder view on the extent of educational inequality in Latin America than
53
See Barro and Lee (1993) and (2001). Their data are derived from the UNESCO Statistical
Yearbooks. By means of a perpetual inventory method enrolment rates are reconfigured into attainment
levels of two samples of the working-age population; ages 15-64 and ages 25-64. In addition Barro and
Lee have calculated the distribution of the working age population over seven categories of attainment
levels. The distribution of the labour force among these categories refers to the highest level attained:
1) no schooling, 2) uncompleted primary schooling 3) completed primary schooling, 4) uncompleted
secondary schooling, 5) completed secondary schooling, 6) uncompleted tertiary schooling and 7)
completed tertiary schooling. Compared to previous cross-country datasets (Kaneko 1987,
Psacharapoulos and Arriagada 1986) the Barro and Lee dataset has been a significant improvement in
terms of coverage and distributional detail. The data are sensitive to the assumptions applied in the
perpetual inventory method used to determine the working age population. De la Fuente and Domenech
(2002) have revised the data to correct for inconsistencies in a sample of OECD countries. However,
these inconsistencies are unlikely to disturb the comparative results of different indicators using the
same dataset.
90
international comparisons of standard deviations or secondary school completion shares. In
fact, each of these three indicators throws a highly distinct, and sometimes even opposing,
light on the extent of educational inequality in countries and regions. Table 4.2 compares the
regional averages of the three indicators in the year 2000 for a sample of 101 countries using
strictly identical attainment figures from the Barro and Lee dataset (Barro and Lee 2001). The
Gini-coefficient (G) of the attainment distribution is defined as,
n
n
∑∑ x
G=
i =1 j =1
− xj
i
* (n / n-1)
2n µ
2
Where xi and xj are the average years of schooling of n consecutive quintiles of the
distribution (so n = 5) and µ is 1/n. The quintile distribution is also used by Castello and
Domenech (2002), whereas Thomas et al. (2001) use a septile distribution. The Ginicoefficient ranges from a minimum value of zero, when all quintiles have attained an equal
proportion of the total years of schooling of the labour force, to a theoretical maximum of
one, if the top quintile has received all education and the rest none. The standard deviation (σ)
of the attainment distribution is defined as,
σ=
1
n
n
∑
i =1
( xi − x )2
where n = 5. The secondary school completion share (SSCS) is defined as,
n
SSCS =
∑
i=1
x
ssc
i
x
i
Where n is the total amount of people in the labour force and xissc refers to the share that has
completed secondary schooling as the highest level attained.
Table 4.2 shows that according to the Gini-coefficient average educational inequality
in 21 LAC’s is at par with the world average. It is lower than in Asia, Sub Saharan Africa and
the Middle East and higher than in Europe and the Western Offshoots and the Transition
Economies. However, the standard deviation places Latin America substantially above world
average and suggests that educational inequality in the region is higher than in Asia and
considerably higher than in Sub Saharan Africa. Finally, according to the share of the labour
91
force that has completed secondary schooling, Latin America (8.2%) is far below Asia and the
world average of 12.3%, though much higher than in Sub Saharan Africa.54
Table 4.2: A regional comparison of educational inequality by three different indicators,
unweighted averages, 2000
No. of
Gini-coefficient
countries
attainment
% completed
Standard
secondary
deviation
schooling
attainment
Latin America
21
0.55
8.5
5.0
Asia
17
0.58
13.7
4.6
Sub Saharan Africa
23
0.73
4.0
3.7
North Africa and Middle East
10
0.63
13.0
5.6
Transition Economies
9
0.31
18.3
4.2
Europe and Western Offshoots
20
0.32
21.7
4.6
World
100
0.54
12.3
4.50
Sources: Authors own calculations based on Barro and Lee (2001).
Note: Countries included are Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Cuba, Ecuador,
El Salvador, Guatemala, Haiti, Honduras, Jamaica, Mexico, Nicaragua, Panama, Paraguay, Peru,
Uruguay, Venezuela (Latin America); Afghanistan, Bangladesh, China, Hong Kong, India, Indonesia,
Japan, Korea rep., Malaysia, Myanmar, Nepal, Pakistan, Philippines, Singapore, Sri Lanka, Taiwan,
Thailand (Asia); Botswana, Cameroon, CAR, Congo Dem. Rep., Congo Rep., The Gambia, Ghana,
Kenya, Lesotho, Liberia, Malawi, Mali, Mozambique, Niger, Rwanda, Senegal, Sierra Leone, South
Africa, Swaziland, Togo, Uganda, Zambia, Zimbabwe (Sub Saharan Africa); Algeria, Egypt, Arab.
Rep., Iran, Iraq, Israel, Jordan, Kuwait, Syria, Tunisia Turkey (North Africa & Middle East); Bulgaria,
Croatia, Czech Rep., Estonia, Hungary, Poland, Romania, Slovakia, Slovenia (European Transition
Economies); Belgium, Canada, Cyprus, Denmark, Finland, France, Greece, Ireland, Italy, Netherlands,
New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, UK, USA (Europe and New World).
Why the Gini-coefficient does not capture what we think it does
To see why these indicators provide such a different view on the comparative extent of
educational inequality in Latin America (as a region) we analyse the three indicators in some
more detail focusing on a temporal comparison between 21 Latin American and 8 East Asian
countries as presented in table 4.3. The table shows that the Latin American Gini was
significantly lower in the early post-war era, that it declined substantially in both regions and
54
The interpretation of educational inequality in Sub Saharan Africa runs into even larger trouble:
according to the Gini-coefficient African countries obtain the highest levels of educational inequality,
while the standard deviation suggests they obtain the most egalitarian levels in the world. The standard
deviation reflects an “absolute” rather than a “relative” spread in years of schooling. Ram’s (1990)
analysis, based on educational attainment data of 100 countries decomposed into 6 categories of
attainment derived from Psacharopoulos and Arriagada (1986), suggests that educational inequality is
subject to an inverted U-curve identical to the Kuznets curve: increasing educational investments first
enhance educational inequality and after a turning point at approximately 7 years of attainment
convergence sets in.54 Ram does not particularly assess the case of Latin America, but the main
conclusion of his empirical analysis is that there is no relation between educational inequality and
income inequality.
92
that this decline has been larger in East Asia than in Latin America. Hence, present levels of
educational inequality, as expressed by the Gini-coeffcient of the attainment distribution, are
significantly higher in Latin America than East Asia, but previous levels were not.
This interpretation is problematic. The Gini-coefficient captures the “relative
variation” in years of attainment, a concept which is extremely sensitive to differences in
“average levels” of education.55 Especially if the distribution contains a share of the labour
force with zero to one year of schooling, the Gini tends to become an almost perfect substitute
for primary school enrolment rates. The correlation coefficient of the Gini and the percentage
share of the working age population without schooling is -0.95. Appendix figure A.4.5
presents a scatter plot of this correlation for 1846 observations in the Barro and Lee dataset. It
shows that the relation is not only extremely tight, but is also subject to heteroskedasticity.
When we remove the category of “no schooling” from the distribution and re-estimate the
Gini, the indicator will remain highly sensitive to levels of attainment in the lower quintiles.
Table 4.3: Regional averages of the Gini-coefficient, standard deviation and coefficient
of variation, Latin America (21 countries) versus East Asia (8 countries), 1950-2000
1950
1960
1970
1980
1990
2000
Gini-coefficient
Latin America
0.71*
East Asia
0.67
0.64
0.60
0.57
0.55
0.77
0.67
0.56
0.51
0.43
Standard Deviation
Latin America
2.72*
East Asia
3.24
3.47
4.09
4.66
4.98
4.03
4.33
4.53
4.92
5.11
Coefficient of Variation
Latin America
East Asia
1.27*
1.17
1.14
1.04
0.97
0.91
1.36
1.13
0.93
0.84
0.69
Source: Authors own calculations based on Barro and Lee 2001.
Notes: Countries included are Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Cuba, Ecuador,
El Salvador, Guatemala, Haiti, Honduras, Jamaica, Mexico, Nicaragua, Panama, Paraguay, Peru,
Uruguay, Venezuela (Latin America); Hong Kong, Indonesia, Korea rep., Malaysia, Philippines,
Singapore, Taiwan, Thailand (Asia); * Excluding Bolivia, Brazil, Cuba, Dominican Rep., Honduras,
Jamaica, Peru and Uruguay. The underlying data are presented in appendix table A.4.4 and A.4.6.
Given the “level-dependency” of the Gini-coefficient of educational distribution, the 1960 and
1970 figures reflect the fact that primary school enrolment rates were higher in Latin America
than in East Asia (see section 3). The East Asian Gini decreased much faster because its
55
The relative difference between 0.5 years of schooling and 1 year of schooling is the same relative
difference between 5 years and 10 years of schooling. Their absolute differences are 0.5 and 5 years of
schooling respectively. This problem also occurs with income Gini’s, but these generally suffer less
from this bias since there is always a substantial amount of income in the bottom brackets of the
income distribution (see Atkinson 1983: pp. 53-6 or Sen 1997: pp. 29-31). With an alternative
comprehensive measure of educational inequality such as the Theil coefficient one encounters the same
problem.
93
average years of attainment rapidly overtook those in Latin America after accomplishing full
primary school enrolment rates, while some LAC’s (i.e. El Salvador, Haiti, Guatemala,
Nicaragua, Bolivia) stayed behind. In sum, the Gini tells us as much about the average
attainment levels as about the extent of variation around this average.
Contrary to the Gini, the standard deviation focuses exclusively on the absolute
variation in attainment years around the average. With a standard deviation of around 5 years
of schooling there is hardly any difference between both regions in the year 2000. Yet, only
the coefficient of variation, which divides the standard deviation by its mean, makes the
amount of variation between populations with different means comparable. The coefficient of
variation shows that, controlled for the differences in the mean of the two regions, the
variation appears to be considerably higher in Latin America in 2000.
This has not always been the case. In the 1960’s the coefficient of variation in East
Asia was higher than in Latin America, but since then it declined much faster which indicates
that the process of educational expansion in East Asia has been more dynamic. Here it should
be pointed out again that it is useful to distinguish between comparative levels of educational
inequality, which presumably peaked at a higher level in East Asia (at 1.36 in 1960), and the
comparative pace of convergence in the attainment distribution. Although Latin America, as a
region, did not reach a coefficient of variation as high as 1.36, it did take much longer before
all children attained a reasonable amount of schooling years. Hence, a relatively large
variation in attainment persisted much longer.
This conclusion may be flawed because of larger intra-regional variation underlying
the Latin American average. Appendix table A.4.6 therefore also presents the rate of change
in both indicators between 1960 and 2000 for each individual country in the sample. These
figures show that only Haiti has witnessed a decline in its coefficient of variation that is close
to the average East Asian country. This is not surprising in the light of the extremely high
initial level of Haiti in 1960. All other LAC’s recorded a considerably slower pace of decline.
Controlling for the initial levels one may compare Peru and Venezuela with Taiwan, or
Nicaragua with South Korea and Singapore. Thailand appears to be an East Asian country
with a rather “Latin American” outlook.
Secondary School Completion rates
Secondary school completion shares of the attainment distribution give an indication of the
size of the educational “middle class”. The inference is that the larger this share, the lower the
level of educational inequality. This interpretation, again, only makes sense controlled for the
level of educational accumulation. It is possible to have a low percentage share of secondary
school completion and a perfectly egalitarian distribution, when all individuals in the
94
population have completed primary school but did not extend their educational career. Figure
4.3 shows secondary school completion shares controlled for the average years of schooling
in Latin America and East Asia. As noted above, the accumulation of schooling years was
more rapid in East Asia than in Latin America. However, controlled for average attainment
levels, the secondary school completion share was substantially higher in East Asia as well.
In sum, the “mean controlled” coefficient of variation and the secondary school
completion share confirm that the traces of educational underdevelopment as revealed in
section 4.3 are still present in the Latin American labour force of the early 21st century. Both
indicators show that the process of convergence of individual attainment levels has been
much slower in Latin America than in East Asia. It should be emphasized that recent changes
in the distribution of education, that is changes in the flow rather than the stock variables,
have only marginally affected the attainment data of the year 2000 (which after all reflects the
working age population from 25 to 64). Hence, it is too early to conclude that the negative
distributional consequences of educational expansion still prevail. In fact, the grade enrolment
distribution approach in the next section raises some support for the view that LAC’s have
started to break away from their historical legacy of educational inequality in the last two
decades of the 20th century.
Figure 4.3: Secondary school completion shares (y-axis) versus average years of
schooling attained, Latin America versus East Asia, 1950-2000
25%
20%
15%
10%
5%
0%
0
2
4
Latin America 1950-2000
6
8
10
East Asia 1960-2000
Source: Figures are retrieved from the dataset of Barro and Lee 2001.
95
4.5 A grade enrolment distribution approach, 1960-2005
This section introduces a new indicator of educational inequality which focuses on levels of
grade repetition and drop out rates in primary and secondary schooling. The methodology has
been developed in papers by Frankema and Bolt (2006) and Frankema (2008). The core idea
of this approach is that the percentage distribution of grade enrolment rates in primary and
secondary schooling contains information on grade repetition rates and pre-completion drop
out rates. The grade enrolment distribution thus provides insight in the effectiveness of
educational systems with respect to extorting regular school attendance and supporting
children in the process of grade promotion towards school completion. Since the data on
grade enrolment rates offer a much larger amount of detail than the “standard” gross or net
school enrolment rates, changes in the distribution of education can be analysed at a more
detailed level.
The percentage distribution of grade enrolment in primary and secondary schooling
can be obtained from UNESCO’s Yearbook of Statistics for five-year intervals from 1950
onwards.56 The grade distributions of primary and secondary schooling can be linked together
using the absolute number of pupils enrolled in both levels of schooling and weighing their
respective percentage distributions according to the following formulas,
Xp
X p + Xs
* g pi
,
Xs
* g si
X p + Xs
Where Xp and Xs refer to the total number of students enrolled in, respectively, primary and
secondary schools and gpi and gsi refer to the percentage share of students enrolled in the ith
grade of primary and secondary school.57
Depending on the total amount of grades in primary and secondary education a
standardised distribution can be obtained for ten to twelve grades for 92 (former) developing
countries and 32 OECD countries from 1960 onwards. Table 4.4 presents two examples of
this standardised grade enrolment distribution in Argentina and Canada for the year 1960. In
the hypothetical scenario that each grade contains exactly the same amount of students, all
twelve grades would contain 100/12 = 8,33%. In practice, the grade distribution is always
skewed towards the lower grades because some children leave school earlier than others.
Most OECD countries reveal a pattern comparable to Canada’s, where the percentage shares
56
From 1999 onwards the data are accessible online (UNESCO Institute for Statistics (UIS)).
In some countries there is an overlap in the final grades of primary and the first grades of secondary
schooling that requires extra calculations to link the series adequately. Generally the students in the
“intermediate” grades were added to the first grades in secondary education.
57
96
decline more rapidly only in the final grades (9 to 12). At this point some children have
(already) completed their secondary school. Developing countries reveal patterns that are
more comparable to Argentina in 1960, or even far more skewed. Assuming, for the moment
(we will discuss the validity of this assumption and possible solutions further below), that the
influx of children in the system is constant, a considerable amount of children either repeats
one or several of the lower grades for one or more years, or drops out before reaching the
higher grades, or both.
The weak record of Latin America regarding grade promotion and school completion
has been widely acknowledged in literature (see for instance Schiefelbein 1992, Martin 1994,
Birdsall et al. 1997). Figure 4.4 illustrates this “stylized fact” by picturing the grade enrolment
distribution in Colombia and South Korea in 1970. It should be noted that in Colombia the
reported gross enrolment rate in 1970 is 102% and in South Korea 104% (UNESCO,
Statistical Yearbook 1974). In other words, practically all children attend primary education in
both countries. Nevertheless, the grade enrolment distribution in Colombia reveals an
enormous contrast compared to South Korea.
Table 4.4: The percentage distribution of grade enrolment in Argentina and Canada in
1960 (12 consecutive grades in primary and secondary schooling)
Grade
1
2
3
4
5
6
7
8
9
10
11
12
Argentina
21.3
14.0
13.8
12.0
10.2
8.7
7.2
4.2
2.9
2.4
1.8
1.4
Canada
11.9
11.1
10.8
10.3
10.0
9.7
9.3
8.4
7.1
5.1
3.8
2.4
Source: UNESCO, Statistical Yearbook 1972, authors own calculations.
Figure 4.4: Percentage distribution of grade enrolment in Colombia and South Korea,
1970
South Korea 1970
25
15
20
12
% of pupils
% of pupils
Columbia 1970
15
10
5
9
6
3
0
1
2
3
4
5 6
grade
7
8
9
10
0
1
2
3
4
5
6
grade
7
8
9
10
Source: UNESCO, Statistical Yearbook 1972 and 1978-1979 (authors own calculations).
Notes: three year moving average of twelve consecutive grades in primary and secondary schooling.
97
In Colombia high rates of grade repetition and pre-completion drop out rates skewed the
grade enrolment distribution towards the lower grades. Only a small group of children
completed primary schooling and enrolled in secondary schooling. Those who did had a
relatively large chance of completing secondary school compared to children in primary
school. On the other hand, Korean children were more evenly distributed among the first six
grades of primary schooling, while the grade distribution in secondary schooling was more
skewed. This simple comparison not only exemplifies the Latin American context of
educational expansion, it also shows the large limitations of gross enrolment rates for
comparative purposes.
Tentative explanations for grade repetition and pre-completion drop out
Irregular school attendance goes a long way in explaining the phenomena of grade repetition
and pre-completion drop out. Children can be officially enrolled (i.e. registered) without
attending in practice. Absenteeism has multiple causes that are more often than not related to
poverty: a lack of finances to cover school expenses, a lack of school transportation and
prohibitive distances to schools in rural areas, overcrowding of schools, health problems of
the child (undernourishment), child labour, a lack of perceived interest of schooling by
parents, a lack of support and attention by teachers, insufficient monitoring on attendance and
performance, and so on and so forth.
The problems of absenteeism and irregular school attendance have been recognized
for a long time in Latin America. In the 1956 report for the Brazilian Institute for Education,
Science and Culture, composed by J.R. Moreira, it is shown that 53.1% of all Brazilian pupils
are enrolled in the first grade, 21.8% in the second, 15.5% in the third and 9.7% in the final
fourth grade. Moreover, 42.7% of the children leave school without ever passing the first
grade and over 70% leaves school before completing four years of education. Out of the other
30% the majority of pupils spent five, six or seven years to finish four grades. The report
states that,
“In a country which is obviously poor in spite of its present extraordinary industrial
development, we fix something which is capable of change and revision, and keep the child in
one primary grade for two, three or more years or even turn him out of school before he has
learnt the least it can give him.” (UNESCO 1958, World Survey of Education II, p. 172)
And with respect to the poor regions in the North East of Brazil the report states,
98
“…retardation in the primary schools reaches alarming proportions, expanding and
enlarging the school age band, multiplying the first grades, crowding the classroom, and
dividing the school periods into two, three, or even four sessions because there are not
enough funds to build more schools.” (UNESCO 1958, World Survey of Education II, p. 172)
Framing the grade enrolment distribution into a comprehensive indicator
The distributive information contained in grade enrolment rates can be standardized for
broader comparative purposes by estimating the likelihood that children entering school will
have a smooth school-career up to completion of either primary or secondary schooling. A
possible method is to take the ratio of the percentage share of students in grade 1 to the shares
in grade 6, 9 or 12. The disadvantage of this approach is that such a comparison is sensitive to
year-to-year fluctuations that occasionally occur in school enrolment. An alternative method
is to average out these fluctuations by taking a weighted measure of students enrolled in more
than one grade. This will also reveal a larger part of the underlying structure of the
distribution. Any ratio of grade enrolment rates is feasible once we normalise the equation for
the number of grades involved as follows,
GDR 1-N =
∑g
∑g
i = ( n +1), N
i =1, n
i
n
N −n
*
i
Where N is the total number of grades and gi is the percentage share of enrolled in the ith
grade. Since the majority of countries have adopted a six grade elementary curriculum a
measure including the first six grades gives the best fit to standardize the inequality indicator
for primary schools. Assuming that the influx of pupils is constant over time, the ratio of the
grades 4 to 6 over 1 to 3 expresses the chance that a pupil in grades 1 to 3 reaches the higher
grades 4 to 6 without repeating grades or dropping out. The GDR 1-6 is defined as,
GDR 1-6 =
∑g
i
∑g
i
i = 4 −6
i =1−3
So far, the implicit assumption has been made that the influx of pupils is constant over time.
A growing (or declining) school-age population skews the grade enrolment distribution, if it
implies that each year more children enroll than in the previous year, other things equal. The
99
countries under consideration here almost all witnessed rapid increases in their school-age
populations (the 5 to 14 year old category) over the period 1960-2005. Demographic growth
generally explains the bulk, between 75 and 100%, of year to year fluctuations in total
enrolment. What are the potential effects on the GDR created by the demographic factor?
The demographic database of the UN provides population figures for the age group 5
to 14 from 1950 onwards (five year intervals, see UN World Population Prospects 2004)).
For three regions and a group of least developed countries the average annual growth rates
has been calculated for each decade. Table 4.5 shows the average annual growth rates for the
entire period 1960-2005 in Africa, Asia, Latin America and the least developed countries. To
estimate the maximum possible impact of demographic change on the GDR 1-6, we also
included Latin America in the decade 1955 to 1965 in the last row: the annual increase of the
Latin American age cohort 5-14 at a rate of 3.4% was the highest being encountered.
Table 4.5: The effects of population growth on the grade distribution, annual growth of
age group 5-14, 1960-2005
Annual
growth
(age 5-14)
grade distribution
1960-2005
1
2
3
4
5
6
GDR 1-6
Distortion
Africa
0.026
11.4
11.1
10.8
10.5
10.3
10.0
0.926
0.074
Asia
0.013
10.7
10.5
10.4
10.3
10.1
10.0
0.962
0.038
Latin America
0.015
10.8
10.6
10.5
10.3
10.2
10.0
0.955
0.045
least developed countries
0.026
11.3
11.1
10.8
10.5
10.3
10.0
0.927
0.073
Latin America (1955-1965)
0.034
11.8
11.4
11.1
10.7
10.3
10.0
0.905
0.095
Sources: Annual population growth figures taken from UN, Population Prospects 2004, medium
variant.
The outcome of this exercise is that, in the extreme case scenario, demographic growth can
distort the GDR by almost 0.10, ceteris paribus, and in some individual cases even slightly
more.58 This potential spatial and temporal bias in the comparison of GDR’s warrants
correction. Fortunately, the demographic data, i.e. the average annual decadal growth rates of
the age cohort 5-14, required for adjusting the original GDR are readily available. So we
obtain the adjusted GDR by:
adjusted GDR xi = original GDR xi + correction xi
58
[4.1]
Given the variation around the Latin American mean (1955-1965). In many OECD countries the
effect of declining birth rates results in a positive, albeit less substantial, bias.
100
where x refers to the country and i to the year of observation. To account for the time-lag
involved in the effect of changes in the influx of students on the GDR 1-6, the annual decadal
growth rates were taken ca. five years in advance of the observation (depending on the exact
year of observation of the original GDR). For example, the observation for South Korea in
1963 and Guatemala in 1961 are both adjusted for the average annual growth rate of the age
cohort 5-14 over the years 1955-1964.
Latin American grade enrolment ratios in international comparative perspective, 1960-2005
Table 4.6 shows the estimates of the grade distribution ratio (GDR 1-6) in the period 19602005 for five developing regions in the world. The first line of each region presents the
unadjusted and unweighted estimates, the second line presents the same estimates, but
weighted according to the total number of students enrolled per country and the third line
presents the weighted and adjusted averages of the GDR 1-6. The underlying data, i.e. the
unadjusted and unweighted estimates, are presented in appendix table A.4.7. The GDR’s
increased around 0.22 to 0.26 between 1960 and 2000 in four of the five regions, but not in
Latin America. In the latter region the increase in the GDR between 1960 and 2000 was 0.42.
It should be noted however that the initial levels of the Latin American GDR in the 1960’s
were staggering low. In other words, the quality of the educational systems that have been
erected during the 20th century in LAC’s, at least in terms of grade promotion and school
completion, was far below the general standard. Part of the rapid increase in the GDR,
therefore, has to be interpreted as a form of “catching-up” convergence.
Table 4.6: Interregional comparison of grade distribution ratios (1-6), weighted and
adjusted averages, 1960-2005
1960/5
1970/5
1980/5
1990/5
2000/5
Latin America (19)
0.42
0.57
0.64
0.68
0.84
South & West Asia (5)
0.50
0.57
0.58
0.73
0.74
East Asia & Pacific (7)
0.63
0.72
0.72
0.93
0.87
Sub Saharan Africa (19)
0.59
0.68
0.79
0.80
0.75
North Africa & Middle East (10)
0.66
0.79
0.81
0.90
0.88
Source: Frankema 2008, UNESCO, Statistical Yearbook, various issues 1962-1998 and UNESCO,
Institute for Statistics, www.uis.unesco.org; Notes: Countries included are Argentina, Bolivia, Brazil,
Colombia, Costa Rica, Chile, Cuba, Ecuador, El Salvador, Guyana, Honduras, Mexico, Nicaragua,
Panama, Paraguay, Peru, Trinidad & Tobago, Uruguay, Venezuela (Latin America); Afghanistan,
Bahrain, Bangladesh, India and Iran (South & West Asia); Hong Kong, Indonesia, South Korea, Laos,
Malaysia, Philippines, Thailand (East Asia & Pacific); Botswana, Burkina Faso, Congo Rep., Ethiopia,
Gabon, The Gambia, Ghana, Ivory Coast, Kenya, Lesotho, Madagascar, Mauritania, Mauritius, Niger,
Nigeria, Rwanda, Senegal, Uganda, Zambia (Sub Saharan Africa); Iraq, Israel, Jordan, Kuwait,
Morocco, Qatar, Saudi Arabia, Syria, Tunisia, Turkey (North Africa & Middle East).
101
There is more evidence supporting this conclusion. Figure 4.5 shows 57 countries that have
achieved full primary enrolment (defined as 95% or above) in the period 1960-2005 and the
adjusted GDR’s 1-6 in the first half of that particular decade. In particular LAC’s turn out to
have combined full enrolment rates with very low GDR’s. For instance, in 1980, Jordan
achieved full gross enrolment rates and complete grade enrolment equalization in the same
decade, whereas Brazil and Nicaragua achieved full gross enrolment rates with an adjusted
GDR of only 0.27 and 0.32 respectively. Or compare Chile in the 1960’s with South Korea or
Singapore, or Colombia in the 1970’s with Zambia, Sri Lanka and Mauritius. All LAC’s
obtained a GDR below 0.75 when achieving full enrolment. The expansion of primary
education in Latin America took place at the expense of the quality of the educational system
and this was a widely shared feature among all LAC’s.
Focusing on the time lag between the achievement of full primary school enrolment
rates and the GDR passing 0.95 we find striking global differences. In Malaysia, Singapore
and Jordan there was no time lag whatsoever, which means that the development of the
educational system not only guaranteed enrolment for all children, but also effectively
organized the system of grade promotion and prevented children to drop out of school before
completion at the same time. South Korea, Cyprus and Mauritius witnessed a one decade
time-lag between reaching both goals. However, after complete primary school enrolment in
the early post-war era it took Argentina five decades and Chile four decades to accomplish
grade enrolment distribution equalisation. Panama and Uruguay are currently approaching a
five decade lag. Hence, the Latin American strategy of educational development can be
characterised as “enrolment over completion”.
102
Figure 4.5: Grade Distribution Ratio (1-6) in the first decade of full primary school
enrolment, 1960-2005
Dominican Rep.
Congo, Rep.
Costa Rica
T urkey
Chile
Uruguay
T unisia
Panama
Philippines
Hong Kong
Korea, Rep.
Cyprus
Singapore
1960's
Colombia
Myanmar
Vietnam
Mexico
Libya
T ogo
Ecuador
Cameroon
Peru
Lesotho
Swaziland
Syria
Qatar
Zambia
Sri Lanka
Mauritius
1970's
Brazil
Nicaragua
Laos
Honduras
Nigeria
Zimbabwe
Madagascar
India
Indonesia
Kuwait
Kenya
T hailand
Iran
Iraq
Botswana
Malaysia
Jordan
1980's
Nepal
Bangladesh
Rwanda
Gabon
Bolivia
Uganda
Algeria
1990's
Guatemala
Benin
El Salvador
T anzania
0.00
2000/5
0.25
0.50
0.75
1.00
Source: UNESCO, Statistical Yearbook, various issues 1966-1999 and UNESCO, Institute for Statistics
(UIS), www.uis.unseco.org. GDR’s from Frankema 2008.
103
Changes in the shapes and slopes of the grade enrolment distribution
Turning back to table 4.6 another stylized fact of Latin American educational development
demands our attention: in the four benchmark regions the GDR’s increased around 0.22 to
0.26 between 1960 and 2000. Yet, in Latin America the increase in the GDR was 0.42. Part of
the rapid increase in the GDR has to be interpreted as a form of “catching-up” convergence,
but it does indicate that improvements in the quality of the educational system were made at
an accelerated pace. In the four benchmark regions the rise of the GDR has recently come to a
halt.59 In East Asia there has been a significant set back and in Sub Saharan Africa the
stagnation has set in already in the 1980’s. Latin America forms an exception. The region
witnessed a temporary slow down during the 1980’s, which has been more than compensated
for during the 1990’s. This implies that current generations of young workers entering the
labour force have received their education in a system that was markedly more equal and,
presumably, offered a higher level of educational quality, than that of their parents. In a more
detailed analysis of the shapes and slopes of the entire grade distribution this conclusion can
be confirmed.
Figure 4.6 presents the grade distribution curves of twelve countries (Argentina,
Brazil, Chile, Colombia, Mexico, Peru and Venezuela, and Kenya, Egypt, India, South Korea
and Malaysia) in 1960 and 2000 (1990 for Brazil and India). The charts include an estimation
of the gradient of the grade distribution curve. The inference is that steeper downward slopes
(i.e. lower coefficients) represent a more skewed grade distribution and higher levels of
educational inequality. The coefficients are reported in the upper right hand corner of each
graph for 1960 and 2000 (or 1990).
There appears to be a great similarity in the shape of Latin American curves around
1960. Convex curves indicate that the grade distribution is highly skewed towards the lower
grades and tend to get flatter in secondary education. Argentina is the single exception to this
pattern in 1960. The majority of benchmark countries reveal an inverted S shape curve
indicating a larger relative emphasis on, or a larger relative success in, supporting children on
the path towards primary school completion. Only the curve of India resembles those of the
LAC’s, albeit with a less pronounced convexity.
The inverted S-shape in Egypt and Kenya has largely remained the same, while the
distribution as a whole has become more equal. Chile reveals a shift away from convexity
59
The recent stagnation can be explained by two factors. First, there has been a setback in some
countries affecting the regional means (Afghanistan, Indonesia, Iraq). Second, a more widespread
slowdown signals decreasing marginal returns on efforts to equalize the grade distribution by means of
supporting school attendance and preventing pre-completion drop-out rates, or it signals a reduced
effort as such. In Sub Saharan Africa the effects of the growth disaster and continuous political
instability since the 1980’s are the most likely explanation for the observed stagnation.
104
towards an inverted S. The other LAC’s seem to have followed the pattern best exemplified
by Peru: the convex curve gradually transforms towards linearity, indicating that there is still
a large bias in the grade distribution towards the lowest grades, yet this bias has become
considerably less pronounced in the four decades before 2000. Finally, in South Korea and
Malaysia the original inverted S curve has now approached the horizontal line that indicates a
perfectly equal distribution of grade enrolments across primary and secondary schooling.
Judged by the changes in the slope of the curve, progress in five of the seven Latin
countries can be considered above average. Argentina, Chile, Colombia, Mexico and Peru all
have witnessed a sharp drop in the coefficient. Mexico stands out with a spectacular drop
from -2.67 to -0.77, but also Argentina and Peru have made respectable progress from -1.83
to -.036 and -1.71 to -0.50 respectively. Compared to for example South Korea (-1.30 to 0.53) the progress can be considered as complete “catching up”, although it should be noted
that Malaysia (-1.60 to -0.17) outperformed all LAC’s except Mexico. Venezuela has lagged
behind somewhat and progress in Brazil has been rather poor (even taking into account the
end year 1990).
The comparative grade enrolment distribution analysis has shown that the expansion of
primary school enrolment rates in 20th century Latin America has, to a large extent, taken
place at the expense of the quality of the educational systems. Levels of grade promotion and
school completion were, controlled for gross enrolment rates, very low compared to other
developing regions in the early post war period. The efforts to repair these shortcomings have
increased during the post-war period. In particular during the 1990’s progress in grade
enrolment equalization was outstanding. This result creates some leeway for the conclusion
that at present LAC’s are rapidly breaking away from a long period of educational inequality
instilled by the neglect of quality maintenance during previous stages of spreading mass
public schooling. Yet, it is too early to witness the effects of grade distribution equalization
trickling down in the labour force.
105
Figure 4.6: The grade enrolment distribution in primary and secondary schooling, Latin
America versus a selection of non Latin American countries, 1960-2000
Chile 1960-2000
Argentina 1960-2000
20
1960 = -1.65
20
1960 = -1.83
2000 = -0.66
2000 = -0.36
16
% of pupils
% o f p u p ils
16
12
8
4
12
8
4
0
0
1
2
3
4
5
6
7
8
9
10
1
2
3
4
grade
5
6
7
8
9
10
grade
Colombia 1960-2000
Mexico 1960-2000
1960 = -2.67
1960 = -1.87
28
2000 = -0.85
28
24
2000 = -0.77
20
% of pupils
% o f p u p ils
24
16
12
8
20
16
12
8
4
4
0
0
1
2
3
4
5
6
7
8
9
1
10
2
3
4
Peru 1960-2000
1960 = -1.71
28
2000 = -0.50
24
% o f p u p ils
16
% of pupils
6
7
8
Brazil 1960-1990
20
12
8
4
9
10
1960 = -1.90
2000 = -1.42
20
16
12
8
4
0
0
1
2
3
4
5
6
grade
106
5
grade
grade
7
8
9
10
1
2
3
4
5
6
grade
7
8
9
10
Venezuela 1960-2000
India 1960-1990
1960 = -1.42
1960 = -1.70
20
20
1990 = -0.99
2000 = -0.95
16
% o f p u p ils
% of pupils
16
12
8
12
8
4
4
0
0
1
2
3
4
5
6
7
8
9
1
10
2
3
4
5
7
8
Korea, Rep. 1960-2000
10
Malaysia 1960-20000
1960 = -1.60
1960 = -1.30
20
20
2000 = -0.53
16
2000 = -0.17
% of pupils
16
12
8
12
8
4
4
0
0
1
2
3
4
5
6
7
8
9
1
10
2
3
4
Egypt 1960-2000
6
7
8
9
10
Kenya 1960-2000
1960 = -2.12
1960 = -1.54
20
5
grade
grade
20
2000 = -1.20
2000 = -1.07
16
16
% o f p u p ils
% of pupils
9
grade
grade
% o f p u p ils
6
12
8
4
12
8
4
0
0
1
2
3
4
5
6
grade
7
8
9 10
1
2
3
4
5
6
7
8
9
10
grade
Source: UNESCO, Statistical Yearbook, various issues. See also appendix table A.4.6 (authors own
calculations).
107
The conclusion that Latin America is breaking with a history of educational inequality may be
true for some, but certainly not for all LAC’s. To illustrative this point figure 4.7 shows the
absolute amounts of public spending per level per student in Argentina and Honduras for four
benchmark years between 1950 and 2000. Maddison’s GDP estimates denoted in 1990 GearyKhamis US dollars were used to convert the percentage share of public educational
expenditure in total GDP into a PPP-adjusted monetary unit (Maddison 2003). Argentina has
reached ratios of public spending per student per level, which are largely comparable to the
majority of OECD countries. In Honduras the ratio of tertiary to primary spending per student
has also been declining since 1954, but absolute public expenses on children in primary
education were extremely low initially. In 1994 public spending per tertiary student was
approximately six times as large, whereas in 1954 this ratio reached nearly twenty. Although
the relative gap has been narrowed, the absolute gap in spending per student per level has
only further increased, from 111 to 239 GK-dollars, which means that a large share of the
extra money available today is spend on tertiary students rather than on children in primary or
secondary school.
Figure 4.7: Total public expenditure per student per level of education in Argentina and
Honduras, 1954-1990 (1990 Geary-Khamis US $)
600
490
500
428
386
400
329
300
200
100
239
228
238
262
175
149
88
84
0
Argentina 1954
108
Argentina 1970
Argentina 1984
Argentina 1991
300
276
262
250
199
200
150
117
100
59
35
34
50
6
15
20
30
37
0
Honduras 1954
Honduras 1970
Honduras 1982
Hondruas 1994
Source: Maddison 2003; UNESCO, World Survey of Education, 1958; UNESCO, Statistical Yearbook,
various issues.
4.6 Conclusion
The present chapter has assessed the development and distribution of mass education in Latin
America from 1870 to 2000 in an international comparative perspective. Special attention has
been paid to the timing and pace of educational expansion and educational equalisation. Since
the distribution of education is inextricably related to the spread of education, comparative
levels of educational inequality (variation) were controlled for the stage of educational
development (accumulation). Although the analysis has indicated a large extent of intraregional variation in educational development throughout the late 19th century and 20th
century, at least three shared Latin features of educational development and distribution can
be recorded:
1) With respect to average levels of GDP per capita, the transition towards mass public
schooling occurred later than in the rest of the New World, Europe and Japan. The start of
three phases of expansion, each referring to a specific set of LAC’s can be dated around ca.
1870, 1920 and 1950. Once underway, the increase in primary school enrolment was not any
slower or faster than could be expected on the basis of the patterns observed in the rest of the
world: it was faster than in the most advanced countries and it was notably slower than in the
poorer developing countries.
109
2) More than in any other part of the world, the expansion of public primary education took
place at the expense of the quality of education. It took even the advanced Southern cone
countries at least four decades to achieve acceptable levels of grade promotion and school
completion after having achieved full primary school enrolment rates. Correcting enrolment
figures for the grade enrolment ratio thus revealed that educational development (and
distribution) significantly lagged behind in Latin America from an international comparative
perspective.
3) In the post-war era levels of educational inequality were gradually reduced. This process
was partly hampered by the economic crises in the 1980’s, but since the start of the 1990’s
Latin America broke away from its traditional path of educational retardation and its
inherently high levels of educational inequality with more speed than witnessed before. The
advances in the reduction of repetition and pre-completion drop out rates were larger than in
other regions of the world. However, large gaps in years of attainment (and the quality of
years attained) can still be observed in Latin America’s labour force at present, as it takes
time before the effects of these improvements trickle down.
These conclusions are based on an analysis of a wide range of educational indicators, which
have not all shed similar lights on the comparative development of Latin American education.
In particular the Gini-coefficient of the attainment distribution provided a much milder view
on the extent of educational inequality in Latin America. The reason for this deviation from
the more general picture is related to the “level-dependency” of this indicator. The advantage
of the use of a wider range of partial indicators of educational inequality is that it offers
various complementary insights into the distribution of education. The grade enrolment
distribution can be considered a useful contribution to the existing set of indicators, since it
nuances the analysis of historical gross primary school enrolment rates in two ways. First, it
shows that enrolment registration differs from actual school attendance. Second, it helps to
differentiate between years of attainment which are usually equally valued: not every
registered year attained has been equally valuable for the student, when taking repetition rates
or pre-completion drop out into account. Since grade enrolment data are available from the
1960’s onwards, the GDR sheds new light on the historical comparative analysis of
educational development and distribution.
The question why LAC’s were so late in improving the quality of their public education
systems has only been tentatively addressed in this chapter, but looking ahead to the second
part of this study, it may be stressed once again that the initial conditions of inequality that
had evolved in the colonial settler societies had a long lasting impact on the comparative
110
development and distribution of education in post-independent Latin America. As long as the
colonial model of the stratified rural society, characterised by high land inequality and various
forms of labour coercion prevailed, a broadly supported expansion of public education was
unfeasible. Land lords needed cheap labour and children of the landed elite were better of
with private education. Given the low prospects of social mobility in these pre-modern rural
societies, the demand for popular education was also limited. Hence, fundamental changes in
government policies regarding mass education largely depended on the decline of the
traditional social order and the political stronghold of the landowning elite. Three factors
played a key role in this process and these will be extensively discussed in the remainder of
this thesis.
1) Structural economic change, primarily urbanisation and industrialisation, altered the
demand for skilled labour and offered new job alternatives to the traditional rural population.
Education became an increasingly valuable asset and a new class of urban entrepreneurs
developed a vested interest in educational expansion.
2) Globalisation, apart from temporarily strengthening the position of the large landowners,
enhanced structural and institutional change in the long run. The forces of the global market
enforced policy makers to reconsider and reform traditional economic policies and it also
induced the spread of new ideologies concerning the position of the poor and the labouring
class.
3) Demographic change was the silent driving force of a dramatic change in the relative
supply of unskilled labour in the course of the 20th century. Whereas the newly independent
Latin American nation states started out with, on average, very low levels of population
density, all of them were at the end of the 20th century characterised by an abundance of
unskilled labour and large rates of underemployment. The traditional labour market
institutions which were designed in response to chronic labour scarcity in the colonial era,
became rapidly obsolete during the 20th century. Consequently, the perceived importance of
education changed in the mind set of policy makers, entrepreneurs and the broader layers of
society.
111
112